In　this　article,　we　consider　variable　selection　in　semiparametric　varying-coefficient　spatial　autoregressive　models　with　a　diverging　number　of　parameters.　With　the　nonparametric　functions　approximated　by　B-spline　basis　functions　and　combining　2SLS　method　with　the　SCAD　penalty,　we　propose　a　variable　selection　procedure.　Under　mild　conditions,　we　establish　the　consistency　and　oracle　property　of　the　resulting　estimators　for　parameter　components　and　consistency　of　the　regularized　estimator　for　nonparametric　component.　Some　simulation　studies　are　conducted　to　assess　the　finite　sample　performance　of　the　proposed　variable　selection　procedure,　and　the　developed　methodology　is　illustrated　by　an　analysis　of　the　Boston　housing　price　data.